Home › esci (Estimation Statistics with Confidence Intervals) › esci in jamovi

esci in jamovi

esci is now available as a module in jamovi!

esci (exploratory software for confidence intervals)

+  jamovi (the free, open-source stats software that crushes SPSS)

a match made in heaven.

Get esci for jamovi:

  • Download and install the current version of jamovi (be sure to select the current version not the solid version), https://www.jamovi.org/download.html
  • Open jamovi, and then under the modules menu (a big + sign), select “jamovi library”

  • The window that appears shows the various modules available for jamovi.  Browse to find esci and then click “Install”

  • If all goes well you’ll end up with an esci menu in jamovi.

What is esci for jamovi?

Geoff has long been working on to make it easy to adopt and use the new statistics.  Back in 2001 he released the first version of ESCI (Exploratory Software for Confidence Intervals), a set of Excel worksheets that provided:

  • Simulations to explore key statistical ideas (e.g. the dance of the means)
  • Analysis pages to provide effect sizes and confidence intervals for most basic study designs (including meta-analysis)
  • Beautiful graphs focused on visualizing uncertainty

Over the years, ESCI has been updated and expanded, with major new releases accompanying the publication of Understanding the New Statistics and the first edition of Introduction to the New Statistics.

esci for jamovi just one part of a big step forward for these efforts.  esci is no longer tied to Excel, but is now being developed as a odule in R that can be plugged in as a module for jamovi .  If that doesn’t mean anything to you, try this:  You can now adopt a new-statistics approach using modern (and free) statistical software.  Estimation just got even easier for all.  Huzzah.

The second edition of Introduction to the New Statistics is currently in preparation. It will make full use of esci, and so will offer everything that used to be accomplished in Excel and more, but now with easy data file handling and natural progression paths to the whole world of statistical software, beyond the intro course.

What can you do with this module?

esci in jamovi now supports most of the basic analyses you would learn in an undergraduate statistics course and meta-analysis (which really should be part of a good undergraduate statistics course). 

Here’s how you would map traditional hypothesis tests onto the analyses available in esci:

Traditional hypothesis test esci in jamovi command
One-sample t-test Estimate Mean
Independent samples t-test Estimate Independent Mean Difference
Paired samples t-test Estimate Paired Mean Difference
One-Way ANOVA Estimate Ind. Group Contrasts
2×2 ANOVA Estimate Ind. 2×2
2×2 Chi Squared Estimate Proportion Difference
Correlation test Estimate Correlation
Correlation test with Categorical Moderator Estimate Correlation Difference

In addition, you’ll find analyses to conduct meta-analysis for 2-group designs (from raw data or Cohen’s d) and for correlational designs, all with the option to analyze categorical moderators. 

Your feedback is needed:

Be warned that esci is still in development; it is still subject to revision and change as we move towards the 2nd edition of ITNS.  Or, maybe instead of feeling warned, you could feel excited, as this means you can have an impact on how this module develops.  We’d be very excited to have your feedback, feature requests, and/or bug reports.  Please especially consider esci through the eyes of your students:

  • What other analyses would you like to see?
  • Anything in the output that is hard to understand?  That should be labelled better?  That should be added or could be removed?
  • Would it be helpful to add the option to see all assumptions for an analysis?  Should we provide more guidance on interpreting output? 
  • Any options missing from analyses?

The best way to provide feedback would be on the github page for this module, which is here: https://github.com/rcalinjageman/esci.  If that’s a hassle, then by all means just email Bob directly or tweet at us @TheNewStats

As you provide feedback, keep in mind a couple of key design goals for this module:

  • For now, esci will only support frequentist statistics.  A long-term goal would be to make it easy to also obtain Bayesian estimates and boostrarpped estimates, but that’s not in the cards for now.  In the meantime, if you want bootstrapped analyses, DaBest already makes this easy: http://www.estimationstats.com/#/. 
  • We want esci to be useful for researchers, but also accessible for students.  So we’ve tried to make this easy to use, have tried to provide lots of feedback and error-checking to prevent mistakes (no more means of gender, please!), and have tried to keep output very straightforward.  We’re still feeling our way here.
  • We’ve tried to keep esci ‘lightweight’, with relatively few dependencies on other R packages. 
  • We’ve separated out the exploration/simulation aspects of esci (like the dance of the means) to focus just on analysis.  We will also be updating the exploration/simulations for the 2nd edition, but it seems likely at this stage that we’ll be making this a separate effort, most likely built for the web rather than in a software package. 

Thank you

There’s still a lot of work to go on this module, but it is probably never too early to be grateful for those who have provided help and assistance.  These would include:

  • Students in Bob’s 2019-2020 statistics and neuroscience classes, who helped put this module through its paces.  Thanks, and may you never have to sideload again.
  • Adam Claridge-Chang and Joses Ho, developers of DABEST.  We benefited tremendously from adopting portions of their source code for esci. 
  • Jonathon Love, Damian Dropmann, and Ravi Selker– the jamovi team.  Jonathon was especially helpful (and patient) with suggesting improvements to the module.

Also in R, you said?

Yes, under the hood of this jamovi module is an R package for estimation.  You can grab it from github if you are interested, but at this point it is not ready for prime-time: code doesn’t follow R style very well, there are no unit tests, and there is no documentation.  We’ve road-tested the module through a 1-year undergrad stats course, so we feel pretty good that it generates appropriate output for the module, but we’re not going to go back to refactor and improve the underlying R package.  Once that’s done, we’ll be posting about it and working to get the package on cran so that esci can be available for R users as well. 

If you’re a JASP user, stay tuned–we’re also exploring possibilities for a module there as well.  The dream is one code base for a well-documented and tested R package that can easily be plugged into a GUI environment.  That will be estimation for all. 

An example:

ManyLabs1 included a replication of Nosek et al. (2002) in which students who identified as male and female were asked to take an implicit association test (IAT) measuring attitudes towards math vs. art.  The key research question is to what extent males and females differ in their implicit attitudes towards math. 

Here’s data for the valid participants fom the OSU lab from ManyLabs 1 (we obtained this data from the OSF site for ManyLabs 1) in .csv format.  You can open this directly in ESCI.

The effect size of interest is the mean difference between two groups, so we will use “Estimate Independent Mean Difference” from the ESCI menu. 

We then specify d_art as the dependent variable (this is the variable showing the IAT score, where scores over 0 represent more positive attitudes towards art, scores less than 0 represent more positive attitudes towards math, and a score of 0 indicates similar attitudes).  For the grouping variable, we enter sex (the gender that participant identifies with).

For output we get a table of means that includes the key effect size: the estimated mean difference with a confidence interval.

We also get a standardized effect size with confidence interval.

And we get this lovely visualization, one which shows all the data, which graphically shows the effect size, and which works to make uncertainty of the estimate salient (hat tip to the DaBest team; we benefited tremendously from adopting portions of their source code)

Other details:

esci is free and open source.

Source code is on github: https://github.com/rcalinjageman/esci

Current road map:

  • Refactor underlying R code for a consistent functions and results objects.  Follow R style throughout and add tests and documentation for all functions.
  • Implement a consistent approach to also providing hypothesis tests, specifically for interval nulls. 
  • Determine a more consistent output and make options for more advanced outputs and/or assumption checks
  • Consider extended analyses for multiple DVs
  • Add more complex designs and improve interface for complex one IV